Summary and effect analysis of methods for calculating binocular parallax

Summary and effect analysis of methods for calculating binocular parallax

This is what I tried in the process of calculating binocular parallax , Want to record . It would be better if it were helpful to everyone ！ If there are mistakes or improvements , Welcome to correct ！！

1. matlab Self contained disparity function (BM Method ) Binocular parallax

disparityRange = [-16 112];
disparityMap = disparity(left,right,‘BlockSize’,…
5,‘DisparityRange’,disparityRange,‘UniquenessThreshold’,3);

The renderings are as follows ：
Original picture 1

Parallax map 1（ Up for BM Algorithm , for SGM Algorithm ）

PS： The effect is OK , On the whole , A large part of the effect presented follows disparityRange Change by change . It is important to choose the appropriate range of parallax . You can start with matlab With APP Choose from Image viewer, Roughly check the parallax range of the two images . An example is shown below ：

Original picture 2

Parallax map 2

Parallax range in the figure above [0,64], The effect is not obvious

Parallax range in the figure above [0,32], The effect is relatively obvious , But it's not very good .
Infer from the second attempt , When the scene in the image is complex ,disparity The calculation accuracy of the function is relatively low .

2. NCC Measure

The renderings are as follows ：
Original picture

Parallax map

3. Block matching to calculate binocular parallax
Original picture ：
Parallax map ：

4.ELAS Method to calculate binocular parallax
Original picture ：

Parallax map ：

Different algorithms , Different effects , But in the calculation of binocular parallax , Parameter setting is particularly important . If the appropriate parameters are not selected , The effect is unspeakable ... alas
It's recorded here for the time being , If you encounter again in the future , Add again ...